COMPARISON OF TWO REAL-TIME IMAGE PROCESSING SYSTEM APPROACHES

摘要

This article introduces two realized systems designed for real-time processing of video data. For both systems the necessary performance is reached by parallelizing data processing. Especially during the iconic pre-processing of images such systems are very advantageous. One of both systems uses the large efficiency of modern current GPUs ("Graphics Processing Unit") by assigning the most substantial processing steps to a graphic card. The associated software system permits a comfortable use of this approach at all. The other system was realized as a cluster of conventional dual-core CPUs. A special programming model supports the generation of image processing applications. After both systems were described in their hardware structure as well as their software architecture, the efficiency, usability, flexibility and the necessary expenditure are compared with the help of easy test tasks in order to permit a quantitative evaluation of possible fields of usage.